systematic-debugging

Follow the four-phase process: read error messages completely, reproduce the issue consistently, check recent changes, then trace the bad value backward through the call stack to its origin. Only form a fix hypothesis after this investigation completes.

Replace setTimeout and sleep calls with condition-based waiting that polls for the actual condition you need, such as an event appearing or state changing. The included waitForEvent helpers poll every 10ms with a timeout, which raised one test suite's pass rate from 60% to 100%.

Stop and return to Phase 1 investigation with the new information rather than stacking more fixes. If three fixes have failed, stop entirely and question whether the underlying architecture is the real problem before attempting a fourth fix.

Never, according to the skill's iron law: no fixes without root cause investigation first. Simple bugs have root causes too, and the process is fast for simple issues while guess-and-check thrashing wastes more time even in emergencies.

Use the included find-polluter.sh bisection script with the polluted path and a test file pattern. It runs each test individually and stops at the first one that creates the unwanted file or directory.